Research map / Graph learning / GNN architectures
Efficient Prediction of SO(3)-Equivariant Hamiltonian Matrices via SO(2) Local Frames
Haiyang Yu, Yuchao Lin, Xuan Zhang, Xiaofeng Qian, Shuiwang Ji
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add SO(2) TP to node updates (ablation) | QHNetV2 · QH9-stable-id | Hamiltonian MAE: 36.50 -> 31.50 | Improved |
| add SO(2) TP to node updates (ablation) | QHNetV2 · QH9-stable-id | eigen energy MAE: 429.17 -> 417.89 | Improved |
| add SO(2) TP to node updates (ablation) | QHNetV2 · QH9-stable-id | wavefunction cosine similarity: 98.18 -> 98.58 | Improved |
| add SO(2) FFNs to off-diagonal updates (ablation) | QHNetV2 without SO(2) TP · QH9-stable-id | Hamiltonian MAE: 49.84 -> 36.50 | Improved |
| add SO(2) FFNs to off-diagonal updates (ablation) | QHNetV2 without SO(2) TP · QH9-stable-id | eigen energy MAE: 823.93 -> 429.17 | Improved |
| add SO(2) FFNs to off-diagonal updates (ablation) | QHNetV2 without SO(2) TP · QH9-stable-id | wavefunction cosine similarity: 97.88 -> 98.18 | Improved |
| replace SO(3) tensor products with SO(2) operations (QHNetV2 vs QHNet) | QHNet · QH9-stable-id | training speed: 1.00x -> 4.34x | Improved |
| use QHNetV2 instead of SPHNet | SPHNet · QH9-stable-id | Hamiltonian MAE: 45.48 -> 31.50 | Improved |
| use QHNetV2 instead of SPHNet | SPHNet · QH9-stable-ood | Hamiltonian MAE: 43.33 -> 22.97 | Improved |
| use QHNetV2 instead of SPHNet | SPHNet · QH9-stable-id | eigen energy MAE: 334.28 -> 417.89 | Hurt |
| use QHNetV2 instead of QHNet on a small training set | QHNet · MD17 Water (500 geometries) | Hamiltonian MAE: 10.36 -> 22.55 | Hurt |
| use QHNetV2 instead of QHNet | QHNet · MD17 Ethanol | Hamiltonian MAE: 20.91 -> 12.05 | Improved |
Closest papers by meaning
- TraceGrad: a Framework Learning Expressive SO(3)-equivariant Non-linear Representations for Electronic-Structure Hamiltonian Prediction (ICML 2025 · similarity 0.90)
- Learning the Electronic Hamiltonian of Large Atomic Structures (ICML 2025 · similarity 0.90)
- Efficient and Scalable Density Functional Theory Hamiltonian Prediction through Adaptive Sparsity (ICML 2025 · similarity 0.90)
- Local-Global Associative Frames for Symmetry-Preserving Crystal Structure Modeling (NeurIPS 2025 · similarity 0.88)
- ELECTRA: A Cartesian Network for 3D Charge Density Prediction with Floating Orbitals (NeurIPS 2025 · similarity 0.87)
- Higher-Rank Irreducible Cartesian Tensors for Equivariant Message Passing (NeurIPS 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
